The bigger they are the less likely they’ll fit. The pore sizes of each hollow fiber is a random distribution. Little things - urea, water - fit through them all and are sensitive to gradient TMP. Larger solutes - protein bound molecules - don’t have as many pores large enough to fit through and aren’t cleared by gradient differences, some bump through a pore that fits but it is more they are dragged across by the shear force along the membrane and convective clearance, or really any time UFR > 1.0 which is not a good idea unless you’re getting replacement fluids as you dialyze, for instance hemodiafiltration.
The shear force is little credited but plays a huge role in getting non-tiny things across the membrane. Each fiber inside the artificial kidney is a tube and as such the fluid traveling down the center of the tube will travel faster than the fluids bumping along the walls. The effect of this is that the density of solutes in the center of tube will be lower than the density along the walls. Think of all the solutes starting down the tube in a line, as the ones in the center move faster the density is decreased, this density imbalance causes molecules to be pulled from the wall to the center to keep the density the same.(atoms can not abide density differences, the same force creates the lift that keep A380s in the air). So now there is an opening on the one side of the membrane.
However, this is happening on both the inside of the tube and the outside. The fluid traveling fastest will have the most atomic-level suction force. TMP puts a thumb on the scale moving solutes into the dialysate, but when Qd > Qb everyone is pulling/pushing in the same direction.